A practical methodology for experimental fault injection to test complex network-based systems

Abstract

A methodology for dependability assessment of complex computer systems, such as fault tolerant grids, is presented in this paper. The methodology uses communication fault injection and was built by adapting a widely accepted approach for performance analysis. To demonstrate its applicability and usefulness, the methodology was applied to a third party grid platform using a fault injector we are developing. The paper reasons about the advantages of this methodology to perform fault injection campaigns in the prototype phase of a system.